Pharma Companies Without AI Capabilities Will No Longer Be Considered Pharma

This article is based on publicly available information and is intended solely for informational exchange. It does not constitute any investment advice.

In 1913, Henry Ford installed the first moving assembly line at his factory in Highland Park, Michigan. Before, it took 12.5 hours to assemble a Model T; afterward, just 93 minutes.

The reactions of peers fell into two categories. Some tore down old workshops and rebuilt them in Ford's image, while others said, "How can such a shoddy product possibly compare with hand-polished craftsmanship?" In less than fifteen years, nearly all car companies that stuck to hand workshops had vanished.

The assembly line not only transformed the automobile industry but also rewrote the underlying logic of manufacturing: before its advent, manufacturing capability rested in the hands of workers; afterward, it resided within the factory's system.

More than a hundred years later, the same logic switch is happening in the pharmaceutical industry:

At 0:00 AM Beijing Time yesterday (May 20), the most memorable part of the 2026 Google I/O conference was not just that Stitch added real-time voice collaboration, not just that Gemini Spark learned from OpenClaw to work 24/7 in the cloud, and not just that Gemini 3.5 Flash outperformed 3.1 Pro by more than 300 points on the Benchmark.

The truly important thing was hidden in the closing segment of the press conference, buried in Demis Hassabis’s almost overlooked statement. He said the goal is to reimagine the drug discovery process until one day curing all diseases becomes a reality (one day solving all disease).

01

A New Pipeline

Over the past three decades, the efficiency of drug research and development has not increased but decreased. There is a self-deprecating saying in the entire industry called "Eroom's Law" —— spelling Moore's Law backwards. Moore's Law states that the performance of chips doubles every 18 months, whereas Eroom's Law says that for every $1 billion invested in R&D, the number of new drugs approved halves every nine years. This is not the failure of a single company. It is a systemic dilemma.

Now, Google is taking this dilemma apart and reassembling it.

The first tool, AlphaFold. AlphaFold 3 not only predicts protein folding, but also simulates the interactions between proteins and DNA, RNA, small molecules, and ions, which are the core mechanisms of how drug molecules work in the body.

In the past, determining the structure of a protein could take months of X-ray crystallography or cryo-electron microscopy experiments, and success was not guaranteed. AlphaFold 3 has compressed this process to just a few hours, essentially rewriting the rules. Hassabis revealed that over the past two years, the amount of protein structure data accumulated by Isomorphic Labs has exceeded the total produced by all traditional experimental methods in human history.

The second tool, Gemini for Science. This research version of Gemini, introduced by Google at the I/O conference, can track the latest papers, convert research objectives into usable code, and generate new hypotheses. It integrates more than 30 life science databases such as UniProt, AlphaFold Database, AlphaGenome API, and InterPro, reducing complex analyses that previously took hours to just a few minutes.

The third tool is a set of AI drug design engines: IsoDDE, which Isomorphic Labs is preparing to fully promote after receiving $2.1 billion in Series B financing (see "AI Drug Discovery Just Secured the Largest Financing of $14.3 Billion").

The logic is clear when the three tools are stacked together: AlphaFold is responsible for "visualizing" the molecular structure of diseases, IsoDDE is responsible for "designing" drug molecules targeting these sites, and Gemini for Science is responsible for generating hypotheses, designing experiments, and analyzing results in collaboration with human researchers.

In other words, Google is building a new assembly line for this industry.

The old production line that is currently being replaced is the longest, most expensive, and highest failure-rate production line in the history of pharmaceuticals. Last year, Goldman Sachs provided a set of figures: from target discovery to drug approval, traditional pharmaceutical companies worldwide take an average of 14 years, with cumulative investments exceeding 1 billion US dollars, and over 90% of candidate drugs ultimately fail after entering clinical trials.

The numbers coming out of the new production line are as follows: AI-designed drug candidates can progress from target discovery to the preclinical stage within 13 to 18 months (traditionally, this process takes at least three years). The Phase I clinical trial success rate of AI-native biotech companies has reached 80% to 90%, nearly double the historical industry average of 50%.

02

Pharmaceutical Containers

In the late 1940s, a standardized iron box called a "container" began to transform global trade. Before this, goods were unloaded from ships to trucks and transported from ports to warehouses entirely by dockworkers manually carrying them. For a standard ton of cargo shipped from the U.S. to Europe, nearly half of the shipping cost was attributed to loading and unloading. The container changed the underlying logic of this process: maritime shipping costs were no longer a decisive variable in trade.

What happened next was unpredictable.

Japanese automakers discovered that the previously high shipping costs, which made it impossible to cover the cost of sourcing parts for the North American market, had plummeted. Toyota then invented just-in-time production – parts were delivered to the production line exactly when needed, minimizing inventory costs to the extreme. Later, Nike and Apple realized that since shipping costs became negligible, factories no longer needed to be close to consumer markets. This marked a shift in global supply chains from "local production" to "lowest-cost production," ushering in a complete era of globalization.

It took 20 years from the advent of containerization to the birth of just-in-time production, and another 20 years from just-in-time production to the restructuring of global supply chains.

This arc is almost parallel to the path that AI pharmaceuticals is currently taking.

AlphaFold has been on the road for four years, much like the stage when containers were mocked at the port as "standardized rough stuff." But it is quietly seeping into every crack of the pharmaceutical chain. AI drug design engines like IsoDDE are more akin to the early Toyota production lines, integrating multiple processes into a coherent system rather than serving as an acceleration tool for any single step.

By 2025, the global market for AI in drug discovery is estimated at approximately $3.1 billion, and by 2026 it is projected to reach around $8.8 billion. Different institutions may have varying forecast figures, but all predictions point in the same direction. Notably, this figure was less than $1 billion in 2019.

03

Encyclopedia, Yellow Pages and Cars

When a new technology begins to systematically reduce industry costs, its impact is never mild.

In 1993, the Encyclopedia Britannica had annual sales of $1.2 billion, possessing the world's largest encyclopedia content library and the most authoritative academic brand. In less than three years, Microsoft upended the market with its free bundled Encarta, and soon after, Wikipedia delivered the final blow through open-source collaboration.

The Encyclopedia Britannica was not defeated by a better encyclopedia product; it was rendered obsolete by a new method of content production and distribution logic. The same story happened to newspapers: classified ads were taken away by Craigslist, financial content was captured by Bloomberg and Reuters, and local news was eroded by Nextdoor. The entire process took fifteen years.

The reason why the pharmaceutical industry has not been "digitized" in the past is that its core assets are not information, but molecules. The structure of molecules does not change with the advent of the Internet. But AI is different. What AI deals with is not information exchange, but knowledge discovery—drug research and development is essentially a knowledge-intensive labor, which is exactly the area where AI excels.

Specifically, in the industrial chain, the transmission path of the impact is becoming clear.

The ones feeling the pressure most directly are enterprises that take "trial-and-error efficiency" as their core competitiveness. Against the backdrop of AI restructuring early-stage R&D, the first-mover advantage based on human trial and error is rapidly fading.

A deeper level of restructuring will occur within traditional CROs. Companies like WuXi AppTec and Pharmaron, which focus on labor-intensive experiments in early-stage R&D outsourcing, are directly facing the capability replacement brought by AI platforms. One of the core values of CROs has been their "large-scale human scientist-driven wet-lab systems," but AlphaFold and IsoDDE are shifting the key aspects of early discovery from "experiment-driven" to "computation-driven."

The most fiercely contested battleground is the "last mile" between AI platforms and MNCs. Isomorphic Labs has already established deep partnerships with Novartis and Eli Lilly. This also reveals Google's deepest intention: to use AI to rebuild the entire value chain of drug discovery, and then offer it as a service to pharmaceutical companies worldwide.

This is no different from what Ford did back in the day—not replacing the automobile, but redefining how cars are manufactured. And in this paradigm shift, as with all technological revolutions, the biggest winners are often not the top players of the old system, but the new players who have mastered the new operating system.

04

Pharmaceutical companies without AI will no longer be called pharmaceutical companies.

When a key technical indicator of an industry climbs from 1% to 50%, it most likely won't "happen gradually." It will accelerate after a certain tipping point and then become irreversible.

Electricity powered less than 10% of machinery in American manufacturing in 1900, but by 1930 this figure had exceeded 80%. Early adopters of electricity simply replaced steam engines with electric motors without changing the layout of the factories themselves. The machines still revolved around the same central drive shaft, with transmission belts hanging from the ceiling, providing only limited efficiency improvements.

The real breakthrough came from a person whose name all textbooks have forgotten: Burton Moore. He proposed that the central drive shaft should be removed, allowing each machine to have its own independent motor. This idea was considered almost heretical at the time, as investing in new independent drives was not only costly but also would disrupt production. However, independent drives enabled factories to no longer be arranged around a single shaft, leading to the birth of the assembly line and eventually triggering a full-scale boom in the automobile, steel, and chemical industries.

The current progress of AI in pharmaceuticals is roughly at the critical point of this "motor shaft replacement."

The CPHI 2026 Annual Report predicts that over 50% of approved drugs in the next decade will involve AI, penetrating the pharmaceutical industry at least twice as fast as the previous wave of internet adoption in retail. While this figure may sound like an optimistic assumption today, the pharmaceutical industry has never before experienced such a dense influx of technological advancements — from AlphaFold to IsoDDE, and from $2.1 billion in funding to more than half of the global Top 10 MNCs beginning to sign contracts with AI drug discovery platforms.

In the first quarter of 2026, within just the first 15 days, multinational pharmaceutical companies have already initiated over nine AI-driven drug discovery collaborations, with a total value exceeding $60 billion. Eli Lilly and NVIDIA announced a $1 billion investment to establish a joint AI drug discovery laboratory. These deals collectively indicate that AI is irreversibly penetrating every link in the pharmaceutical chain.

This logic of penetration follows the same pattern as electricity: first emerging as an alternative energy source to improve the efficiency of individual processes, much like how AlphaFold accelerates structural biology today; then gradually evolving into an independent power source that begins to reorganize the entire production process—as Isomorphic Labs is doing by AI-izing the entire chain from target discovery to clinical candidates—and eventually becoming the default infrastructure.

The pharmaceutical industry in 2026 is at the inflection point of this curve: just as a factory without electricity 100 years ago would no longer be called a factory, in the future, a pharmaceutical company without AI capabilities will no longer be called a pharmaceutical company.

       Title: Pharmaceutical companies without AI capabilities will no longer be called pharmaceutical companies